GDB / README.md
jaepoong's picture
Add image-6 benchmark dataset with embedded assets
31dde18 verified
|
Raw
History Blame
6.2 kB
---
license: apache-2.0
task_categories:
- visual-question-answering
- image-to-text
- text-to-image
language:
- en
tags:
- benchmark
- design
- multimodal
- graphic-design
- svg
- typography
- layout
- animation
- lottie
pretty_name: "GDB: GraphicDesignBench"
size_categories:
- 1K<n<10K
configs:
- config_name: category-1
data_files:
- split: train
path: "category-1/train-*"
- config_name: category-2
data_files:
- split: train
path: "category-2/train-*"
- config_name: image-6
data_files:
- split: train
path: "image-6/train-*"
- config_name: layout-1
data_files:
- split: train
path: "layout-1/train-*"
- config_name: layout-2
data_files:
- split: train
path: "layout-2/train-*"
- config_name: layout-3
data_files:
- split: train
path: "layout-3/train-*"
- config_name: layout-4
data_files:
- split: train
path: "layout-4/train-*"
- config_name: layout-5
data_files:
- split: train
path: "layout-5/train-*"
- config_name: layout-6
data_files:
- split: train
path: "layout-6/train-*"
- config_name: layout-7
data_files:
- split: train
path: "layout-7/train-*"
- config_name: layout-8
data_files:
- split: train
path: "layout-8/train-*"
- config_name: lottie-1
data_files:
- split: train
path: "lottie-1/train-*"
- config_name: lottie-2
data_files:
- split: train
path: "lottie-2/train-*"
- config_name: svg-1
data_files:
- split: train
path: "svg-1/train-*"
- config_name: svg-2
data_files:
- split: train
path: "svg-2/train-*"
- config_name: svg-3
data_files:
- split: train
path: "svg-3/train-*"
- config_name: svg-4
data_files:
- split: train
path: "svg-4/train-*"
- config_name: svg-5
data_files:
- split: train
path: "svg-5/train-*"
- config_name: svg-6
data_files:
- split: train
path: "svg-6/train-*"
- config_name: svg-7
data_files:
- split: train
path: "svg-7/train-*"
- config_name: svg-8
data_files:
- split: train
path: "svg-8/train-*"
- config_name: template-1
data_files:
- split: train
path: "template-1/train-*"
- config_name: template-2
data_files:
- split: train
path: "template-2/train-*"
- config_name: template-3
data_files:
- split: train
path: "template-3/train-*"
- config_name: template-4
data_files:
- split: train
path: "template-4/train-*"
- config_name: template-5
data_files:
- split: train
path: "template-5/train-*"
- config_name: temporal-1
data_files:
- split: train
path: "temporal-1/train-*"
- config_name: temporal-2
data_files:
- split: train
path: "temporal-2/train-*"
- config_name: temporal-3
data_files:
- split: train
path: "temporal-3/train-*"
- config_name: temporal-4
data_files:
- split: train
path: "temporal-4/train-*"
- config_name: temporal-5
data_files:
- split: train
path: "temporal-5/train-*"
- config_name: temporal-6
data_files:
- split: train
path: "temporal-6/train-*"
- config_name: typography-1
data_files:
- split: train
path: "typography-1/train-*"
- config_name: typography-2
data_files:
- split: train
path: "typography-2/train-*"
- config_name: typography-3
data_files:
- split: train
path: "typography-3/train-*"
- config_name: typography-4
data_files:
- split: train
path: "typography-4/train-*"
- config_name: typography-5
data_files:
- split: train
path: "typography-5/train-*"
- config_name: typography-6
data_files:
- split: train
path: "typography-6/train-*"
- config_name: typography-7
data_files:
- split: train
path: "typography-7/train-*"
- config_name: typography-8
data_files:
- split: train
path: "typography-8/train-*"
---
# GDB: GraphicDesignBench
40 benchmarks for evaluating vision-language models on graphic design tasks — layout, typography, SVG, template matching, animation. Built on 1,148 real design layouts from the [Lica dataset](https://lica.world).
**Paper:** [arXiv:2604.04192](https://arxiv.org/abs/2604.04192) &nbsp;|&nbsp; **Code:** [github.com/lica-world/GDB](https://github.com/lica-world/GDB) &nbsp;|&nbsp; **Blog:** [lica.world](https://lica.world/blog/gdb-real-world-benchmark-for-graphic-design)
## Usage
```python
from datasets import load_dataset
ds = load_dataset("lica-world/GDB", "svg-1")
```
## Schema
| Field | Type | Description |
|-------|------|-------------|
| `sample_id` | string | Sample identifier |
| `benchmark_id` | string | e.g. `svg-1`, `typography-3` |
| `domain` | string | layout, typography, svg, template, temporal, category, lottie |
| `task_type` | string | `understanding` or `generation` |
| `prompt` | string | Evaluation prompt |
| `ground_truth` | string | Expected answer (JSON for complex types) |
| `image` | Image | Input image (when applicable) |
| `input_image_asset` | Image | Auxiliary source image for generation/editing tasks |
| `mask_asset` | Image | Auxiliary edit mask for generation/editing tasks |
| `ground_truth_image_asset` | Image | Auxiliary target/reference image for generation metrics |
| `metadata` | string | Task-specific fields as JSON |
## Evaluation
```bash
pip install git+https://github.com/lica-world/GDB.git
```
```python
from gdb.registry import BenchmarkRegistry
registry = BenchmarkRegistry()
registry.discover()
bench = registry.get("svg-1")
scores = bench.evaluate(predictions, ground_truth)
```
## Citation
```bibtex
@article{gdb2026,
title={GDB: A Real-World Benchmark for Graphic Design},
author={Deganutti, Adrienne and Hirsch, Elad and Zhu, Haonan and Seol, Jaejung and Mehta, Purvanshi},
journal={arXiv preprint arXiv:2604.04192},
year={2026}
}
```
Apache 2.0